Your stack already has agents, catalogs, pipeline monitors, and metric definitions. Nodal doesn't replace any of them — it sits above them and answers the one question none of them do: is the number right, and will it stay right?
Bring the agent you already picked — or swap it next quarter. Nodal feeds it one governed context over MCP and tests its answers against your ground truth, so a model change is a benchmark run, not a leap of faith. Agent-agnostic means no lock-in: the truth stays yours even when the agent changes. Watch Claude Code and Codex land on the same governed answer →
A catalog describes what data exists. Nodal reads from your catalog as one context source — then measures whether answers built on it actually come back right, and how much each piece of documentation earns in answer quality. Complementary, not competing.
Pipeline observability catches freshness and volume anomalies — the loud failures. Silent SQL is the quiet one: the pipeline is green and the answer is wrong. Nodal regression-tests questions against verified results, so drift in meaning gets caught the way drift in data does.
Nodal's commercial evaluation layer does not require you to replace your semantic or business-context layer. It evaluates whichever context the agent can access, measures its contribution through with-context and without-context runs, and re-runs the suite when that context changes. If you do not have a governed business-context layer, Nodal Context is the free, open-source starting point.
Nodal is the control layer across the tools you already use—not another place to ask a data question.
| Layer | Primary job | What it validates | Relationship to Nodal |
|---|---|---|---|
| AI agent | Interprets a question and produces an answer | Its own execution path | Nodal evaluates the context the agent uses and the resulting answer |
| Catalog / semantic layer | Documents data and defines shared metrics | Metadata and metric definitions | Nodal measures whether those definitions improve answers and whether agents honor them |
| Data observability | Detects pipeline, freshness, and volume failures | Data health and delivery | Nodal catches semantic drift when pipelines remain green |
| Nodal | Evaluates AI analytics across every agent and change | Whether the business answer is right and stays right | The evaluation and observability layer across the stack |
Observability on every question and regression evals across the agents and context you already use. Bring your context layer—or build one with free, open-source Nodal Context. How it works →